Research And Implementation Of Microseismic Recognition And Time Pick-up System Based On Deep Learning

With the development of digital signal processing technology and computer technology,microseismic monitoring system gradually adopts digital signal processing and computer technology for data acquisition,analysis and processing,thus improving the monitoring accuracy and efficiency.In the microseismic monitoring system,the primary problem is the microseismic identification and time pick up,whose accuracy is of great significance to the source location and underground early warning.The research of microseismic monitoring system mainly focuses on microseismic.Microseismic has such problems as high frequency of time series data,fast wave speed and short duration,and vibration characteristics are not easy to be found.Moreover,microseismic signals collected by monitoring system will contain a large amount of noise,which increases the difficulty of microseismic event identification and pick-up.Therefore,in order to solve these problems,this paper proposes the research of microseismic identification and time pick-up methods based on deep learning,and develops software to realize the system functions.After analyzing a large number of domestic and foreign literature,this paper proposes the CNNDet microseismic detection model and the CGANet arrival time picking model for the microseismic detection and arrival time picking problems in the microseismic monitoring system.CNNDet is an event detection model based on convolutional neural networks,which can extract useful features from multiple seismic signals and judge the occurrence of events through a classifier.The model combines the data augmentation method of deep learning to train high-precision models with a small amount of data sets.CGANet is a model that combines gated recurrent units and self-attention mechanisms to accurately pick up the P-wave arrival times of detected events.The gated recurrent units can effectively process sequential data,and the selfattention mechanism can learn to automatically focus on and highlight important features in the model,thereby improving the model’s accuracy.This paper conducts experiments from multiple angles and aspects and tests different signal-to-noise ratio samples.The models can still maintain high accuracy under low signal-to-noise ratio.In actual seismic source location,this method also demonstrates superior performance,providing a new idea for accurately identifying mining microseismic monitoring and disaster such as rockburst.The microseismic identification and arrival time pickup system uses Python,MySQL and other related languages as the main development languages.Firstly,a detailed requirement analysis is carried out for the system,and the functional division is made for microseismic detection and arrival time picking.Then,the detailed design of all functional modules of the system and the system database design are completed separately,and finally,the microseismic monitoring module is implemented and the above event detection and arrival time picking modules are integrated into the functional module.The overall design adopts an object-oriented approach to develop the system,which can make the system more flexible and extensible to adapt to future changes in requirements.